Even modern circuits can demand a lot of current in gulps so a 'reservoir' cap can still be needed.
Of course. This discussion is about *where* those reservoirs belong (based on the earlier article).
Let's say, for example, your circuit which has a switched relay on the board.
Your PCB track line to the 'top' of the coil might be thin and a few cm long.
When that relay (coil) is triggered it will demand a current along that track.
Now you have IZ losses and radiation from that track.
Imagine you put a capacitor across the 'top' of the coil down to ground.
Now, when the coil is switched it can provide most of the energy to energise the coil.
This means a reduced pulse of current down the track as it has formed an RC.
(This can also apply with IRLED transmitter circuits).
And the best place to put the 'reservoir' is as close to the relay/IRLED as you can get it, not some random location between the device and the regulator.
Size?
Depends on the application and how long it needs to provide energy or hold the power up.
And there are ways to calculate the size of a capacitor to support the need of a given device. But the problem is the recommendation to put reservoirs 'somewhere convenient' along the line. How do you calculate the required value of a capacitor that is two inches from the regulator and three inches from the device that needs the reservoir? And if you've provided for the regulator at the regulator and provided for the device at the device, why put something random in between? That is the practise that was not mentioned in the article and yet is done so frequently.
Have you heard of Time Constants?
Is that a serious question or sarcasm?
Where is it needed?
As locally as possible to the thing.
Exactly! That is what I'm saying. And it should not be at some random point along the way.
An engineer from TI wrote an excellent and amusing article about the habit of novices "throwing" capacitors at circuits without proper consideration (I've done it myself before). You often don't need to chuck a zillion microfarads randomly into a circuit. And sometimes it can be harmful.
I have to add this is where simulators can help the grey matter
Again, exactly! And potentially lethal to the circuit if the amount of capacitance exceeds the regulator's ability to handle short circuits or turn-off pulses because there are no protection diodes in the circuit.
Summary.
No, 'reservoir capacitors' are here to stay where needed.
You, as the designer, need to work out the best position and value.
Though, quite often, guesswork and a little over-engineering works just fine
Yes, to all of that. But if best practise is to provide capacitance at the regulator for stability and the rest of the capacitance as close as possible to the device causing the abrupt demand changes, why recommend unspecified (and uncalculable) amounts of capacitance at some random point in the circuit between the regulator and the device drawing current?
And there is always the issue of accidentally exceeding the manufacturer-specified maximum capacitance on the output of the regulator which leaves the circuit susceptible to regulator failure due to backflow of current in some circumstances.
Several people have defended the practise of random reservoirs between supply and device but always using the example of a device suddenly shifting current demand, and this is exactly the situation that demands capacitance local to the device, not some distance down the line. There seems to be no reason to put capacitance in the circuit in locations other than at the regulator and at the devices drawing current, and only in the correct amount.
We've all done it -- capacitance in the line somewhere 'just because' but maybe with consideration we might change our pracises to putting the correct amount of capacitance exactly where it belongs?